The Agent Skills Directory

[COMMAND_EXECUTION]: Utilizes curl to interact with OAuth and OpenID Connect discovery and token endpoints to probe for vulnerabilities.
[COMMAND_EXECUTION]: Includes an embedded Python script template for performing concurrent authorization code redemption attempts (race conditions).
[DATA_EXFILTRATION]: Contains explicit instructions for redirecting authorization codes and tokens to external domains controlled by the tester as part of the assessment flow.
[PROMPT_INJECTION]: Ingests untrusted data from remote endpoints (OIDC discovery documents and browser responses) while maintaining high-privilege capabilities.
Ingestion points: OIDC/OAuth discovery endpoints (/.well-known/openid-configuration) and data extracted from browser interactions via browser_evaluate.
Boundary markers: No explicit delimiters or warnings are used to separate untrusted web content from the agent's core instructions.
Capability inventory: Access to system commands via curl, browser automation, and file system writes for evidence logging.
Sanitization: The skill lacks mechanisms to sanitize or validate content retrieved from the target application before processing.

oauth-attacks